Robotic environments for training and assessing the human motor system

By virtue of their precise sensing and actuation, robotic devices are a possible means to enhance human motor learning in health or following injury, as well as to objectively assess performance, but this potential is still largely unexplored. The overall goal of this dissertation was to develop robotic environments that can be used to improve the understanding, training, and assessment of human arm movement. Three novel environments were developed.; The first environment allowed study of how sensitive the human motor system is to small force perturbations in the presence of relatively large background forces of a different type, during the task of reaching. Experimental studies with this environment demonstrated that the human motor system is able to adapt to a small force field which is superimposed on a much larger background force, and that human motor adaptation operates independently of conscious perception of the small force field, since perception is masked by the large background forces. Further, current computational models of human motor adaptation need to be modified to take into account nonlinearities associated with the background force. A conceptual framework is proposed for modeling how the human motor system blends internal model formation, impedance control, path planning, and force minimization across a range of background force conditions. It is proposed that motor system minimizes a cost function consisting of kinematic error, force, and change of force with the combination of these strategies.; The second environment was designed to help people learn to make a novel movement trajectory, an essential process during human development (e.g. learning to reach, eat, or write during childhood), a training method for stroke patients (e.g. adjusting abnormal movement pattern), as well as a common goal in a wide variety of advanced human endeavors (e.g. surgery, athletics, and fine arts). Two paradigms were compared: using a robotic device to physically guide the patient's arm along a desired, novel path, and using the robotic device to visually demonstrate the path. Both techniques were effective in helping human subjects learn the novel path. Surprisingly, however, visual demonstration was marginally better than physical guidance. This result indicates the importance of the visual system for human motor learning, and the potential for robots to teach humans by visual demonstration.; The third environment developed is an advanced, robotic platform for assessing and training human arm movement. The device allows application of a large range of background forces using pneumatic control and force feedback, and can assist in training of complex, naturalistic movements. A software library of robotic assessment techniques was developed to identify factors (e.g. strength, synergy, range of motion, tone, spasticity, impaired sensation, adaptation ability) that limit motor learning following a neurologic injury such as a stroke.